And what makes someone well-suited to do it?
Most people hear about forensic musicologists when they’re reading a news story about a famous musician being sued for stealing another musician’s song. Fairly reliably, the plaintiff’s musicologist testifies they stole it; the defendant’s says they didn’t.
A musician, music supervisor, or advertising producer, however, is more likely to hear about forensic musicologists when the concern emerges that a musical composition might infringe upon another. Someone involved says, “We should get a musicologist on this and see if we have a problem.” And then someone else says, “How do we find a musicologist? And what will they do?”
I don’t think Musicologize has ever really explained all of that. So let’s explain it now.
- What is a forensic musicologist?
- What does a forensic musicologist do?
- And what skills and knowledge does a forensic musicologist need to possess? What makes someone well-suited to being one?
What is a forensic musicologist?
Forensic musicologists are specialists who apply musical expertise, most often involving the analysis of melody, harmony, rhythm, form, and orchestration, to illuminate the relevant truths in matters of civil law.
That’s sorta the “book answer.” I wrote it. Google likes it. That settles it.
What do forensic musicologists do?
Forensic musicologists examine the question, “Given two works, “Song A” and a more recent “Song B,” is there similarity that is likely to have resulted from copying and not merely by coincidence or “independent creation?” And are the elements shared by these works sufficient, and unique or original to Song A, such that this appropriation is unlawful?
And we will get to the skills and knowledge part after we more fully examine what a forensic musicologist actually does.
That was the short version. Here’s the long one:
Forensic musicology analyzes music to answer questions relevant to copyright infringement. What then are the relevant types of questions? Here at Musicologize, music copyright infringement is always described as being a bit of a three-legged stool. The three legs of copyright infringement are:
- Access — “Did the author of the later-created work (Song B) have access to the earlier one (Song A) prior to creating Song B such that they might’ve copied Song B?
- Similarity –“Is Song B similar enough to Song A to be considered a copy, and is it also significant, substantial, or ‘enough’ to matter?”
- Originality — “Does the author of Song A really have a clean title to the aspects Song B perhaps copies? Are the similar elements novel or are they common? Can we find them in prior art? Might the author of Song B have gotten that idea from anywhere?”
Why are those the most important questions a musicologist examines?
The importance of access in copyright infringement.
Infringement requires copying, and copying requires access. Absorb that. To infringe, you must copy Song A, and to copy Song A, you need to have first heard Song A. Sounding the same in the absence of copying and access would be a coincidence. And coincidence is not an infringement. Moreover, as Musicologize has discussed, “access” increasingly looks like the most important of the three legs on the stool. It’s observably the first battle you fight if you’re the defendant.
Access would be easy to establish if defendants simply admitted they had heard the plaintiff’s song. But that never happens. The court, therefore, is forced to consider whether access can be inferred and there are only a few roads:
- Is there a story about how the song made its way directly to the infringer? Let’s call it the “I gave their manager my demo” story.
- Is there a story about how the infringer had sufficient opportunity to have heard it because it’s well-known and widely available? Does it have 300,000 views on youtube?
- Does the music itself tell us access happened? A forensic musicologist’s report can infer access when the similarities are so striking that access can be assumed. This brings us to…
The importance of similarity in copyright infringement.
Similarity is the still the centerpiece of infringement, or non-infringement.
Ostensibly there’s a process that begins with access, then moves to extrinsic similarity (expert analysis) and then to intrinsic similarity (layperson impression), but I’d argue this linear happy path is fantasy. In practice it is all very intertwined — infringement requires copying which requires access which usually requires substantial similarity which is greatly a matter of originality. It’s no exaggeration to say forensic musicology answers all the relevant questions nearly all the time.
Start with Access, which as we’ve said most often must be inferred and therefore a function of substantial similarity such that a musicologist will assert, “there’s no other way these two works could be so similiar!” Substantial similarity meanwhile is distinguishable from insubstantial similarity mostly by virtue of dissectible protected and unprotected elements of ideas and expressions in the music. This is the purview of expert musicologists who are able to factor in (some would say “filter out”) the originality and protectability of the musical elements when evaluating the quantity and quality of the similarity between two works.
The importance of originality in copyright infringement.
Originality informs both the forensics regarding musical similarity and the legal protectability by copyright. A plaintiff must first have a registered copyright to initiate a litigation, and that copyright protects only the original elements in the work. Unoriginal elements might be “building blocks,” the basic elements and conventions in music, available to all authors; scales, chords, sequences, cadences, forms, and so forth. And unoriginal elements obviously include that which can be observed in prior art, things therefore created by another author before the plaintiff. The plaintiff has no claim over such elements. Forensically, the defendant author of Song B cannot be assumed to have copied an unoriginal or commonplace element from Song A or likely from anywhere. And indeed, that which is commonplace is not “probative of copying,” but instead supports a thesis of independent creation.
The importance of and the problem with, Arnstein.
For all the complexities inherent in being hired either by the plaintiff or the defendant, the goal of the forensic musicologist should be truth, and serving justice. Our three-legged stool model best explains the forensic musicologist’s train of thought because access, similarity, and originality are the matters the musicologist can illuminate. The courts however think of copyright infringement as two-pronged; step one and step two, in that order, horse and cart. This approach has problems worth our exploration.
First, the two prongs. Ever since Arnstein, the two prongs have been “copying” and “unlawful appropriation.” And who’s Arnstein? You’ve heard of Cole Porter, right? In the first half of the 1900s, he wrote some of the greatest songs of all time. But unless you’re an intellectual property attorney or a forensic musicologist you have not heard of Ira Arnstein. Arnstein sued Porter back in the day (he sued a lot of people) and Arnstein v. Porter, established the precedent that those two prongs are the two relevant questions to be answered in a copyright infringement case: First, was it copied? And if so, second, does the copying amount to improper or unlawful appropriation? They’re meant to be sequential. Question A is followed by question B. In other words, “Arnstein” wants to know, first, if Songwriter B committed the act, consciously or unconsciously, of copying from Songwriter A. When the answer is yes, we move on to ask if the elements that Songwriter B copied are “enough,” and if they belonged to Songwriter A such that Songwriter A has a claim.
The intended clarity of sequentiality in Arnstein’s two-pronged approach is problematic with regard to musicology because similarity is as integral to the first as to the second of Arnstein’s prongs which, musicologically, are not clear and sequential, but circular and redundant and murky. To boot, the latter prong is the province of the layperson jury, and explicitly not the expert.
The idea that courts ascertain first if copying occurred is fine if the defendant confesses or if you can track the provenance of the proverbial “mp3 demo file the plaintiff gave to someone in the pop star’s entourage.” Those occurrences are rare. Most of the time the courts require inference of access from a combination of a: wide dissemination of Song A (a lot of views on youtube) and b: musical similarity so striking as to preclude coincidence, or “probative of copying.” The latter of which is dependent obviously on everything the second prong should involve. Substantiality of similarity is quantitative and qualitative. It requires originality.
The idea that we might satisfy the copying requirement and then separately whether the copied material constitutes “unlawful appropriation” is fraught in nearly all practical applications of forensic musicologist analysis because a musicology-reliant test A would not be satisfied without the similarities being substantial enough to pass test B. To finally put it succinctly, “probative of copying” includes substantiality which requires originality.”
What does a forensic musicologist actually do?
As we said originally in the short version, musicologists seek to answer these questions:
Given two works, “Song A” and a more recent “Song B,” is there similarity that is likely to have resulted from copying and not merely by coincidence or “independent creation?” And are the shared elements sufficient, and unique or original to Song A, such that the appropriation is unlawful?
“We look at the notes!” That’s certainly the cliche. I say it, and I’ve heard other musicologists say it. The truth is that the process is very involved, and can go in as many directions as any investigation might, but let’s at least cover the basics.
First, the law requires that our methodologies be generally accepted. We can’t employ astrology to figure out if you copied that Taylor Swift song.
Critical listening and transcription.
We listen. We identify the parts of the works that are relevant to our basic questions and we “zoom in” on those parts, figuratively and literally. We transcribe the parts which is to say we write out the music, place notes on staves, and create other graphical visual representations, of which there are several. It would be descriptive if figurative to say we write the music on vellum paper, hold the vellum up to the light, and see if the notes line up. What we actually do is pretty similar. We put the transcriptions on different staves, lined up, time-corrected, and spaced (we use notation software that does the lining up stuff for us) so we can see (and also demonstrate) similar events more clearly. And at this point, quantitatively, we could make observations such as the percentages of notes with “pitch” in common, occurring at like or similar time values, and held for like or similar durations. We might even ask the most obvious question, “How many notes are common to two works?” But while this could be initially impressive, especially to non-experts (judges and juries included), the individual notes aren’t really “music.” In fact, weak arguments that take the form, “29 percent of notes in these 16 measures are the same, either in pitch or rhythmic placement” and argue the probability mathematics, or conversely tout the familiar idea that “there are only 12 notes” are among the coarsest type of analysis.
Musicologists are actually looking for melodies, sequences or strings of notes in the context with their harmonic accompaniment, melodic arcs, instrumental or orchestrational features, the similarity of form, lyrics if there are any, conventions, and unconventionality, and anything we can bring to bear in the service of getting closer to truth.
Everything to this point has identified the quantitative “gross similarities.” We want to eventually get ourselves to an assessment of “net similarities.” Net of what, you ask. Net of what’s musicologically insignificant; things that aren’t original to our “Song A” either by virtue of appearing in prior art, or because they’re excluded from protection by copyright law, elements such as scales.
With gross similarity laid out, the musicologist can analyze more qualitatively. Beyond the sheer “amount of similarity” we might discern, “are these likely or unlikely similarities?” Because, of course, the more unlikely they are, the more they substantiate the unlikeliness of coincidence: the more “probative of copying” the similarities are. This is certainly the stuff of musical expertise itself. Arriving at meaningful values for likeliness or unlikeliness is complex. Lots of data needs to be considered and knowledge brought to bear to arrive at such a judgment and we can’t summarize music theory and history here.
But here’s an example: In my musicological analysis, I might identify “notes that are of like or similar pitch.” I know, it sounds like I’m making stuff up already. “How can pitches be different but similar,” you might ask. Well, they can. Similarity even between two pitches is non-binary. It’s more useful to think of the similarity between two pitches as being on a scale or gradient from most to least similar and there might even be a z-axis that depicts the nature of the similarity. Two things are important here. First, although two musicologists might argue over that scale or gradient, such arguments are vital because they serve justice. And second, despite all this complexity I keep espousing, I could show you, for example, such non-binary similarity, if I have you in the room and you brought your ears with you. Musicologists know and can demonstrate these factors, taking what’s esoteric and making it understandable.
Beyond looking at the notes, musicologists conduct what’s called “prior art research” to ascertain whether the elements in focus can be found elsewhere in music that came before. For one thing, if the similar material was not unique to Song A, regardless of whether it was copied, the owner of Song A doesn’t have a claim to it. Since most forensic musicology work is pre-litigation or even preventative in nature, prior art informs everything we do. While there are lots of tools that help with prior art research, the process is time-consuming and therefore expensive, so an already-at-hand expansive knowledge of music history, on a note-by-note level, is vital to effective and economical musicological analysis.
Prior art considerations in musicological analysis is a science that looks a bit like an art form. There’s an idea, based on legal precedent, that as a part of an analysis of two works, the musicologist should strip out or filter away all that which is unoriginal, in other words, anything that can be found in prior art, and then work with the distillate. In my experience the idea has become a tool wielded by defense counsel and by wrong-minded or even disingenuous musicologists to persuade non-experts. Music, meanwhile, is a fluid medium comprised of contextual elements that sacrifice some, most, or all of their meaning when considered as isolated fragments apart from their whole. The thoughtful musicologist should avoid lazy application of such ideas. Prior art is a consideration, along with unprotectable building blocks in music such as scales, that works best when measured thoughtfully to inform originality, which informs substantial similarity, which informs copying.
Sample detection and analysis.
Most forensic work involves not the recording but the underlying composition, the notes. Sometimes though, musicologists are asked whether one recording contains a sample of another. “Samples” are audio snippets from other recordings, though sometimes the term gets used casually to refer to taking a portion of a composition as though analogous to plagiarism or interpolation. Here’s a whole article on that. Forensic audio analysis on some level is quite a different process, though there are overlaps with compositional analysis. The technological approach might involve acquiring the relevant audio, placing it into sound editing software (which is distinct from digital audio workstations and notation software) identifying the key areas and events in the recording and then the investigation can take you in lots of directions. The musicologist can for example time and pitch-correct them to get a more apples to apples listening experience, in some cases it can be a simple matter of seeing if two audio files “phase cancel” when set against each other (Identical audio will do this. It also happens when you don’t pay attention to the + and – when you wire up your speakers at home — you don’t get strong bass for one thing.) Musicologists can look at the sound waves to see if the waves are identical or if events seem to happen at identical times. Spectral analysis looks at spectrograms which are detailed visual depictions of sound. There’s lots of software especially for that too.
But more in keeping with the analysis musicologists do for composition, we critically listen for conspicuous elements, errors, and artifacts — things that appear or shouldn’t appear in both examples except as a result of sampling. We might hypothesize thoughtfully about how Song B was created and this leads us to drill down and run experiments that validate those hypotheses. An interesting and amusing example: sometimes we might be analyzing two very similar bits of audio, but we find in the course of our qualifying experiments that there’s an element missing in one. What could explain that? We go through the possibilities, using our knowledge of music engineering and production. Sometimes the conclusion we reach is that what we’re hearing as a sample is actually a re-recorded version where an artist might have liked to sample another record, but decided to avoid having to get the license from the record company that owns the masters and just recreated the sample using his own musicians. Sample recreation is relatively common. (You still need a license by the way, usually just from the songwriter.)
Deliverables of forensic musicology?
The goal of the process is to provide an opinion and evidence substantiating that opinion in a manner understandable by our client and perhaps by a legal expert, about the likelihood that song B was copied from song A and that the elements that the two works shared might be unique or original enough to enjoy protection by copyright. In the case of a potential litigant, this might lead to dissuading a client from initiating a claim. Litigation is expensive and difficult. It needs to be well-reasoned and worthwhile. In the case of preventative forensic musicology, the purpose is, prior to publication usually, to avoid infringement or even the accusation of copying, reasonable or unreasonable. There the musicologist assesses that risk, and where concerning risk is found, compositional or production changes can be recommended to mitigate that risk. We change the notes.
What three main skills and knowledge enable a forensic musicologist to answer those questions and provide the service?
The skills and knowledge of a forensic musicologist could be likened to a three-legged stool as well, and those three legs would be…
- Expertise in Music (a three-legged stool by itself!)
- Production Techniques and Technology
- Familiarity with Copyright Law and its History, though you needn’t be an attorney.
- An ability and desire to teach, so you can effectively share your findings with non-expert clients, juries, and judges.
A good and effective forensic musicologist needs all those skills and traits, and here’s why:
First and foremost, you need to be a music expert. You must be able to analyze music quickly, which involves having a great ear for notes and harmony, and the ability to relate it to music theory. You also need a wealth of music history knowledge so that you can relate one work to another, seeing patterns, devices, conventions, novelties, and be able to contextualize what’s in front of you with what has come before. All of this informs answering the forensic musicologist’s basic questions around similarity and originality.
It also helps to know about current music trends and technology, because modern production techniques are often forensically significant. At some point, what really matters is, of course, the music, regardless of how it was composed, but by keeping abreast of technology and modern production techniques, you maintain a more well-informed sense of probability, tantamount to intuition, that produces more efficient hypothesises and serves in truth-finding analysis. Additionally, there are facets of forensic musicology that are inherently technical, such as spectral analysis.
And while you don’t need to be an attorney, you do need a working knowledge of this area of law, how intellectual property law is devised and how it’s litigated. There are a lot of terminologies, precedent, and convention, both musical and legal, that informs how you do the job and serve your client. Precedent matters and past is prologue. Precedents, though, aren’t sacrosanct.
It also helps to have what we hear described as “a heart of a teacher,” which means a natural desire, inclination, and ability to teach. The job doesn’t at all end with the forensic analysis. Forensic musicologists serve people, and most people enjoy music, and have feelings and opinions about music, but are not music experts. Forensic musicology expertise is only valuable when it is shared with non-experts, whether they’re the client or a judge. And that’s a skill in itself. The musicologist needs to be able to teach and persuade for the good of the client, and for the sake of what’s good and just.